On-Chip Etchless and Tunable Silicon Nitride Waveguide Mode Converter Based on Low-Loss Phase Change Material

Abstract

The development of reconfigurable photonic integrated circuits (PICs) demands photonic devices with high-efficiency tuning capabilities, yet conventional thermo-optic and electro-optic methods suffer from limited index modulation and excessive power consumption. To overcome these limitations, we propose an etchless and tunable silicon nitride waveguide mode converter based on low-loss phase change material, antimony triselenide (Sb₂Se₃). By depositing an Sb₂Se₃ layer on the silicon nitride wafer and using a laser-induced phase transition technique, we can write and erase the waveguide structure in the phase change wafer without waveguide etching, where the input/output waveguide is a strip waveguide and the conversion region is built using a tilted subwavelength grating structure. From the results, the obtained TE₀-TE₁ mode conversion efficiency, crosstalk, and insertion loss are higher than 96%, lower than −16 dB, and lower than 0.4 dB at a wavelength of 1.55 µm, respectively. The proposed device enables post-fabrication tuning of the grating duty cycle, allowing working wavelength adjustment for the same device. Furthermore, the device exhibits scalability to other higher-order mode conversions (e.g., TE₀-TE₂). Consequently, we expect that such devices could have important uses in programmable and multifunctional PICs.

1. Introduction

For on-chip optical communication systems, enhancing information transmission capacity is essential for current and future communication technology development [1,2,3]. To address this issue, on-chip multiplexing technologies have emerged as an effective means [4,5,6,7], for example, polarization-division multiplexing (PDM), wavelength-division multiplexing (WDM), and mode-division multiplexing (MDM) [8,9,10]. PDM is fundamentally constrained by its two-dimensional multiplexing capacity and WDM requires costly multi-wavelength laser arrays [11,12]. By contrast, MDM offers a scalable solution by exploiting the orthogonality of eigenmodes in multimode waveguides, which has been rigorously validated through coupled-mode theory and experimental demonstrations [13,14,15]. By multiplexing higher-order waveguide modes (e.g., TE₁, TE₂, TE₃ …), MDM enables a 10-fold enhancement in transmission capacity compared to the single-mode system [16], where high-performance mode converters are the key components. These devices, functioning as higher-order mode generators or multimode optical sources, are indispensable for realizing the on-chip multimode optical transmission or multiplexing transmission. Therefore, new higher-order mode converters with high performance and new functions are worthy of being developed to promote the development of on-chip multimode photonics and programmable PICs [17].

The most common mode converters are based on the asymmetric directional coupler (ADC) and Mach–Zehnder interferometer (MZI) waveguide [18,19,20,21]. The ADC-based mode converter exploits the phase-matching condition between coupled waveguides to enable mode conversion, yet suffers from stringent fabrication tolerance (±5 nm dimensional deviation) and narrow operational bandwidth (Δλ < 10 nm) [18,19]. The MZI-based mode converter, relying on mode splitting, phase difference accumulation, and interference recombination, achieves broader spectral operation at the expense of extended device lengths (>100 μm) [21]. Recent advancements in single-waveguide architectures have demonstrated compact footprints (<30 μm) with high conversion efficiencies [22,23,24]. However, the functions of these aforementioned devices remain unchanged once they are determined or fabricated, which means one device can only achieve one type of mode conversion function. In order to promote the development of on-chip multimode photonics and programmable PICs, we should develop a multifunctional or tunable mode converter, where the mode conversion function could be easily changed on the same device.

In this paper, we propose an on-chip etchless and tunable silicon nitride waveguide mode converter based on the low-loss phase change material, antimony triselenide (Sb₂Se₃) [25,26], which can convert from input TE₀ to output TE₁ mode, and device working wavelength can also be adjusted in real time. The device pattern is formed on the phase change wafer using a laser-induced phase transition technique, where the input/output is a strip waveguide and the key mode conversion region is built using a tilted subwavelength grating (SWG) structure. It should be noted that the photothermal phase transition is achieved through the utilization of laser irradiation, thereby facilitating the phase transition from the amorphous state to the crystalline state of Sb₂Se₃, and we use the refractive index difference (Δn~0.7) to create the etchless waveguide structure. Further, the tilted SWG is used to generate a refractive index gradient that progressively modifies the wavefront of the propagating TE₀ mode and finally we achieve the TE₀-TE₁ conversion with mode conversion efficiency (CE) > 96%, crosstalk (CT) < −16 dB, and insertion loss (IL) < 0.4 dB at λ = 1.55 µm. Meanwhile, the proposed device can achieve the working wavelength adjustment via tuning the duty cycle of the tilted SWG in real time through laser-induced phase transition, which cannot be realized using current lithography and etching facilities. Other higher-order mode conversions can also be achieved based on the proposed device scheme. We hope this device and fabrication technique can provide a new way to boost the development of multimode photonics and programmable PICs.

2. Device Structure and Principle

As illustrated in Figure 1, the schematic of our proposed silicon nitride waveguide mode converter comprises three constituent parts: the input waveguide, the mode conversion region, and the output waveguide. The conversion waveguide region is built using tilted SWG structure and the refractive index difference between the amorphous state and crystalline state of Sb₂Se₃ is used to form the waveguide and grating structure. The width, tilt angle, period, duty cycle, and grating number of the tilted SWG structure are represented by W, θ, Λ, a/Λ, and n, respectively, as shown in Figure 1. The thickness h of the deposited Sb₂Se₃ layer is 80 nm, the thickness H₁ of the silicon nitride layer is 400 nm, and the thickness H₂ of the silicon dioxide layer is 2 µm. Owing to a different waveguide structure compared with previously reported optical waveguides, we offer a detailed process revealing the composition of the etchless waveguide structure, as shown in Figure 2. First, we start from a silicon nitride wafter, and then a film of Sb₂Se₃ layer is deposited using ion beam sputtering or magnetron sputtering. So, the phase change wafer is generated and we can then directly write our designed structure (e.g., input/output waveguide and mode conversion structure in here) pattern on it using the laser processing technique. The deposited Sb₂Se₃ layer is in the amorphous state at first and the laser irradiation technique is used to make the focal point transition to the crystalline state. Further, by combining the movement of the motion stage, the processing laser can write out the designed structural pattern with the crystalline state in the Sb₂Se₃ layer, which is the red region shown in Figure 2.

The fabrication of common optical waveguide structure is dependent on the lithography and etching processes, and the formed strip structure can well confine the optical mode for transmission. By contrast, our proposed waveguide structure does not require a waveguide etching process and uses the method of laser irradiation to generate the refractive index difference in the Sb₂Se₃ layer. Further, with the help of the bottom silicon nitride wafer without any etching process, we can obtain the typical optical waveguide modes (e.g., TE₀, TE₁), as shown in the inset of Figure 1. Therefore, such optical waveguide modes will eliminate the scattering loss caused by the roughness of waveguide etching, contributing to the low-loss transmission on-chip.

For the device working principle, it is based on the refractive index perturbation and phase difference accumulation. When Sb₂Se₃ operates in the crystalline state, the real part of its complex refractive index is 4.05 at a wavelength of 1550 nm. Conversely, when Sb₂Se₃ operates in the amorphous state, the real part of its complex refractive index is 3.28 at the same wavelength, where its imaginary part is close to zero for both states [25]. Using this refractive index difference, we could create the optical waveguide structure and further introduce the refractive index perturbation along the light propagation direction in order to generate the mode conversion. As the input TE₀ mode injects into the device shown in Figure 1, the tilted SWG perturbation structure will introduce a wavefront gradient and one beam of input light will become two light beams gradually. By optimizing the parameters of the tilted SWG structure, we should make the accumulated phase difference of the two generated light beams equal to π and then the new TE₁ mode will be generated. Moreover, based on our proposed laser-induced phase transition technique, we can easily change the structural parameters in real time and this feature will enable us to make structural modifications even after the device has been fabricated, which cannot be realized by a common optical waveguide structure using current lithography and etching processes [27]. Meanwhile, the laser-induced phase transition technique is more flexible than the electrically heating method since the latter requires the fabrication of electrodes within the phase change region and the device function will be determined and cannot be changed after the electrodes are fabricated [28]. Therefore, we use the laser-induced phase transition technique to write our designed mode conversion structure on the phase change wafer and study its online tunable features.

3. Results and Discussion

Before we conduct the structural parameter analysis, we provide the definitions of device performance indicators first. Here, the mode conversion is from input TE₀ to output TE₁ mode. Mode CE is defined as [29,30]

$$\mathrm{C}\mathrm{E}={\displaystyle \frac{P_{{TE}_1}}{P_{out}}}\times 100\%,$$

(1)

where P_TE1 and P_out represent the receiving power of the TE₁ mode and the total power at the device output port, respectively. Mode CT is defined as [29,30]

$$\mathrm{C}\mathrm{T}=\mathit{max}\left(10{\mathit{log}}_{10}{\displaystyle \frac{P_{OT}}{P_{{TE}_1}}}\right),$$

(2)

where P_OT represents the receiving power of the additional modes that interfere with TE₁ at the device output port. The designation mode CT represents the maximum value attained by various modes. In addition, IL is defined as [29,30]

$$\mathrm{I}\mathrm{L}=-10{\mathit{log}}_{10}{\displaystyle \frac{P_{{TE}_1}}{P_{in}}},$$

(3)

where P_in represents the launch power of the TE₀ mode into the device. The three-dimensional finite-difference time-domain (3D-FDTD) method is employed to calculate the device performance [31] and we will use it to find the optimal structural parameters.

For the present device, the tilted SWG structure is pivotal to the mode conversion and its structural parameters should be analyzed in detail, including waveguide width W, grating tilt angle θ, grating period Λ, duty cycle a/Λ, and grating number n. Figure 3 shows the mode CE, CT, and IL of the device as a function of the waveguide width W. The variation in IL is relatively smaller than that of mode CE and CT within the calculation range from W = 1.3 µm to W = 2.0 µm. The allowable width variation range is from W = 1.52 µm to W = 1.82 µm by keeping CE > 90% and CT < −10 dB, where the optimal waveguide width is W = 1.62 µm. So, we should well control the waveguide width during device fabrication and keep its width within the tolerance range. Under the following analyses, the waveguide width W is set as 1.62 µm and we will use it to further obtain the optimum parameters of tilted SWG structure. Figure 4 shows the mode CE, CT, and IL of the device as functions of the grating tilt angle θ, grating period Λ, duty cycle a/Λ, and grating number n. By detailed calculations, their optimum values are θ = 4.5°, Λ = 3.6 µm, a/Λ = 0.45, and n = 4, respectively, corresponding to the mode CE = 97.1%, CT = −16.7 dB, and IL= 0.38 dB. For the parameter tolerance, we also set the same criteria of CE > 90% and CT < −10 dB and the obtained ranges of parameter variations are [4.1°, 4.9°], [3.2 µm, 3.8 µm], and [0.40, 0.55] for the grating tilt angle, grating period, and duty cycle, respectively, where these ranges are marked by light blue-shaded regions in Figure 4. According to these analyses, the key structural parameters of the tilted SWG can be determined and the mode conversion from input TE₀ to output TE₁ mode can be achieved with relatively good performance.

We also conduct the wavelength dependence analysis of the proposed device, where the material dispersion characteristics have already been considered. Figure 5 illustrates the mode CE, CT, and IL of the device as a function of the working wavelength, where the wavelength range is from 1400 nm to 1700 nm. Here, we take the CE value down by 5% compared with the highest CE (~97%) as a criterion. The allowable working wavelength is from 1490 nm to 1670 nm, covering a bandwidth of 180 nm, which would be larger than those mode converters based on ADC and MZI structures [18,20]. Meanwhile, the corresponding CT is less than −13 dB within this bandwidth. For IL, it has quite low wavelength dependence within the whole calculation range and the working wavelength of the proposed device can cover most optical communication bands (S~U), which would help to build on-chip broadband communications.

Figure 6 shows the electric field evolution of our proposed mode converter and the working wavelength is 1550 nm. As the TE₀ mode is injected into the input waveguide along the direction of light propagation, the TE₀ mode was squeezed to one side of the waveguide and then one beam of the TE₀ mode was divided into two beams. After the accumulation of phase difference between these two beams through mode propagation, a new TE₁ mode would be generated at the output port when the phase difference between these two beams is equal to π. Our introduced tilted SWG structure ensured the realization of these aforementioned functions and the total conversion length is 33 μm.

Table 1. Structural parameters of TE₀-TE₁ mode converter.

Structural Parameter	W (µm)	θ (°)	Λ (µm)	a/Λ	n	L (µm)
Optimum value	1.62	4.5	3.5	0.45	4	33

lists the full structural parameters of our designed TE₀-TE₁ mode converter.

Based on our designed TE₀-TE₁ mode converter, we further explore its tunable features with the help of the laser-induced phase transition technique. For an optical waveguide or device fabricated via lithography and etching methods, its structural parameters cannot be changed and the device functions are fixed after fabrication. Thermo-optic and electro-optic methods can only slightly modify the device performance, while the device structure remains unchanged. Here, our used laser-induced phase transition technique can break these limitations and we can directly write and erase arbitrary structural patterns on the phase change wafer, where these writing and erasing processes are real-time and the phase change wafer can be reused repeatedly. The entire process does not require any waveguide etching, which could significantly reduce scattering loss introduced by waveguide etching. For the present device, all the structural parameters of the key tilted SWG can be modified online using the laser-induced phase transition technique. By considering the ease of operation, we choose to modify the duty cycle a/Λ of our proposed device after its structural parameters are determined. Figure 7 illustrates a schematic of the fabrication process for modifying the duty cycle a/Λ, where such modification is achieved through the movement of the motion stage. For example, as the amorphous region of the tilted SWG gradually transforms into the crystalline region using the focused laser on the surface of Sb₂Se₃ film combined with the movement of the motion stage, the corresponding duty cycle of the tilted SWG will increase.

Figure 8 shows the performance of our proposed mode converter (mode CE, CT, and IL) as a function of its duty cycle a/Λ, where a/Λ is varied from 0.35 to 0.6. We can clearly observe that the working wavelength is shifting from long wavelength to short length, revealing a blue shift. The optimum working wavelengths are 1666 nm, 1629 nm, 1568 nm, 1540 nm, 1516 nm, and 1508 nm, respectively, as shown by the dotted lines in Figure 8a–f. If we set the reduction in CE value to 5% as a criterion, the allowable working wavelength ranges are [1583 nm, 1700 nm], [1535 nm, 1700 nm], [1492 nm, 1667 nm], [1470 nm, 1625 nm], [1450 nm, 1598 nm], and [1424 nm, 1581 nm] for the duty cycle a/Λ = 0.35, 0.4, 0.45, 0.5, 0.55, and 0.6, respectively, as shown in the light blue-shaded regions in Figure 8. According to these results, the working wavelength of our proposed mode converter can be easily tuned only by changing its duty cycle even after it is fabricated. This tunable ability and tunable range cannot be realized using previously reported devices fabricated by lithography and etching processes. So, this tunable function is the unique feature of our device, which would be very helpful for on-chip programmable PICs.

We further explore the scalability of our proposed mode converter, which was required for the multimode photonics and on-chip mode-division multiplexing transmission. Some previous reports mainly focused on realizing efficient TE₀-TE₁ mode conversion [32,33,34,35], and few works paid attention to the functional scalability of the mode converter. Here, based on our proposed tunable TE₀-TE₁ mode converter, we further design an on-chip efficient TE₀-TE₂ mode converter, which could convert the input TE₁ mode to output TE₂ mode. Figure 9 plots the electric field evolution of our designed TE₀-TE₂ mode converter based on the tilted SWG structure, and the working wavelength is also 1550 nm. According to the electric field evolution, the working principle of the TE₀-TE₂ mode converter is similar to that of the TE₀-TE₁ one. When the input TE₀ mode entered into the mode conversion region, the TE₀ mode was still squeezed to one side of the waveguide and then one beam of the TE₀ mode was gradually divided into three beams due to enlarged waveguide width. After the accumulation of phase difference between these three beams via mode propagation, a new TE₂ mode would be generated at the device output port as the phase difference between adjacent beams for these three beams is equal to π. The key refractive index perturbation and wavefront gradient was introduced by the tilted SWG structure with optimized structural parameters for generating the new TE₂ mode. The structural parameters of the TE₀-TE₂ mode converter are listed in

Table 2. Structural parameters of TE₀-TE₂ mode converter.

Structural Parameter	W (µm)	θ (°)	Λ (µm)	a/Λ	n	L (µm)
Optimum value	2.5	5.5	1.9	0.4	8	38.6

and the total conversion length is 38.6 µm, where the waveguide width should be enlarged to 2.5 µm. Owing to the excellent tunable features of our proposed scheme, we can also change the device structural parameters of this TE₀-TE₂ mode converter based on the laser-induced phase transition technique and the corresponding device performance, working wavelength range, and device function would also be altered according to different application requirements. Moreover, other higher-order mode conversions with tunable features could be further studied under the guidance of our proposed device scheme and related online fabrication method.

To make comparisons with some previous reported mode converters, we also made a table regarding the device function, length, mode CE, CT, IL, bandwidth and tunable feature of the mode converter, as listed in

Table 3. Device comparisons between previous reported mode converters and present one.

Structure	Function	Length (µm)	CE (%)	CT (dB)	IL (dB)	BW (nm)	Tunable Working Wavelength
ADC [19]	TE0-TM0 [E]	44	>92	<−15	<1	40 (CE > 92%)	No
MZI [20]	TE0-TE1 [E]	~60	-	<−20	~2	-	No
HPS [22]	TE0-TM1 [S]	7	94.6	-	<2.34	35 (CE > 92.2%)	No
PM [23]	TE0-TE1 [E]	4	<90	<−10	1.7	40 (CT < −14.9 dB)	No
	TM0-TM1 [E]	-	-	-	1.2	40 (CT < −10.1 dB)	No
THP [24]	TE0-TM1 [S]	11	-	<−25	4.2	100 (CT < −20 dB)	No
This work	TE0-TE1 [S]	33	97.1	−16.7	0.38	180 (CE > 92%)	Yes

HPS: hybrid plasmonic slot; PM: planar metasurface; THP: tapered hybrid plasmonic; E: experiment; S: simulation; “-”: not mentioned.

. We can clearly find that the wavelength bandwidth and tunable working wavelength are the advantages of the present device, together with its low IL and high CE, where such tunable feature cannot be achieved using current reported device scheme based on the lithography and etching processes. Moreover, the proposed TE₀-TE₁ mode converter can easily expanded to TE₀-TE₂ mode converter only by enlarging the waveguide width and optimizing the structural parameters of tilted SWG structure, revealing good scalability. With these excellent features, we hope our proposed device scheme and related mode converters could find potential applications in the current and future on-chip programmable PICs.

4. Conclusions

In summary, by depositing the phase change material Sb₂Se₃ layer on the silicon nitride wafer, we establish an etchless phase change processing platform and further we design a mode converter based on the tilted SWG structure on this platform, which can achieve the mode conversion from input TE₀ mode to output TE₁ mode. Using the laser-induced phase transition technique, the proposed device pattern is directly written on the phase change wafer and the crystalline state (amorphous state) of Sb₂Se₃ represents the high refractive index region (low refractive index region), where the lithography and etching processes are not required. According to the results, the obtained mode CE, CT, and IL of our proposed TE₀-TE₁ mode converter are >96%, <−16 dB, and <0.4 dB at λ = 1.55 µm, respectively. The working wavelength of our proposed device can be further altered after fabrication only by changing the grating duty cycle with the help of the laser-induced phase transition technique. Moreover, we also explore the scalability of our proposed device scheme and demonstrate the feasibility of the TE₀-TE₂ mode converter. With these features, we hope our proposed etchless phase change processing platform and related mode converters can find potential applications in on-chip programmable PICs.